How LLMs Work (High Level)

Large Language Models are sophisticated prediction engines.

The core mechanism:
1. Training — Model reads billions of text examples from the internet
2. Learning patterns — It learns statistical relationships between words and concepts
3. Next-token prediction — Given some text, it predicts the most likely next word
4. Parameters — Modern models have billions of mathematical weights that encode these patterns

Key insight: LLMs don't "know" things—they've learned what text typically follows what other text. This is why they can write fluently but still make factual errors.